A 0.18-μm CMOS Fully Integrated 0.7-6 GHz PLL-Based Complex
Dielectric Spectroscopy System
Osama Elhadidy, Sherif Shakib, Keith Krenek, Samuel Palermo, and Kamran Entesari
Analog and Mixed-Signal Center, Texas A&M University, College Station, TX, USA
ohatem@neo.tamu.edu, spalermo@ece.tamu.edu, kentesar@ece.tamu.edu
Abstract — A fully-integrated sensing system utilizes a ring
oscillator-based phase locked loop (PLL) for wideband
complex dielectric spectroscopy of materials under test
(MUT). Characterization of both real and imaginary MUT
permittivity is achieved with frequency-shift measurements
between a sensing oscillator, with a frequency that varies
with MUT-induced changes in capacitance and conductance
of a delay-cell load, and an amplitude-locked loop (ALL)controlled MUT-insensitive reference oscillator. Fabricated in
0.18-µm CMOS, the 0.7-6 GHz spectroscopy system occupies
2
6.25 mm area and achieves 3.7% maximum permittivity
error.
138 μm
10 μm
26 μm
56 μm
26 μm
t1
978-1-4799-3286-3/14/$31.00 ©2014 IEEE
4 μm
Passivation Opening
t2
Top Metal Layer
(a)
Δ G MUT (ω ) = ωε r" (ω ) C air
t1
Gsub
I. INTRODUCTION
Broadband dielectric spectroscopy (BDS) is the study of
the frequency profile of a material’s complex relative
[1]. The use of BDS for
permittivity
biosensing applications at radio and microwave
frequencies has been demonstrated, e.g. [2] uses a time
domain reflectometry variant of the BDS technique to
probe biological samples on cellular and molecular length
scales. BDS is also a valuable technique for industrial
applications in material characterization at radio and
microwave frequencies, e.g. detection of concentration,
bulk density, structure, moisture content, etc. [3].
However, the instruments used in the mentioned
biomedical and industrial applications are bulky and
expensive, while also requiring large sample sizes. The
potential system cost and size reduction motivates the
development of fully integrated BDS systems in CMOS
for point-of-care medical diagnosis platforms and for labon-chip industrial sensors.
Previously, CMOS BDS systems based on
voltagecontrolled oscillators (VCO) [4], [5] have enabled
efficient embedding of lumped sensing capacitor elements
within the excitation VCO/synthesizer to allow for selfsustained measurement of the MUT-induced frequency
shift with simple and highly precise readout circuits.
Unfortunately, due to the
-oscillator implementation
these systems can only characterize the real part of
permittivity ( ) over a limited frequency range. Recent
advances in dielectric spectroscopy systems have enabled
complex permittivity detection over extended frequency
ranges. The work of [6] measures MUT-induced changes
10 μm
Csub
Cair + ΔCMUT
= (1 + ε r' (ω ))Cair
(b)
Fig. 1.
model.
(a) Sensing capacitor layout, and (b) its single-ended
in insertion loss of off-chip coupled transmission lines
using parallel low- and high-bandwidth RF modules to
extend detection over a MHz to GHz range. Integration
of on-chip sensors and RF receiver front-ends is achieved
in [7], [8] to enable both real and imaginary ( , )
permittivity detection over wide bands. However, these
CMOS implementations lack the integration of critical
components, such as the sensor [6], frequency synthesizer
(excitation source) and read-out ADC [6]-[8],
necessitating the use of expensive, bulky equipment (e.g.
RF signal generator) for system operation.
This work presents a self-sustained, fully-integrated
PLL-based BDS system which utilizes two ring
oscillators, one loaded with an on-chip sensing capacitor
and the other serving as a reference, for precise
permittivity measurement over a broad frequency range. A
measurement procedure is proposed with a parallel
amplitude locked loop (ALL) that equalizes the sensor and
reference oscillators’ delay cell conductance to enable
quantitative detection of both real and imaginary
permittivity components over 0.7-6 GHz with only a simple
digital counter for frequency shift measurements.
II. VCO-BASED COMPLEX PERMITTIVITY SENSOR
Fig. 1 shows the sensing capacitor, implemented in the
÷
Fig. 2. 3-stage sensing and reference ring oscillators’ delay cells.
top metal layer of a 0.18-µm CMOS process. Exposing the
capacitor to a MUT changes its capacitance Δ MUT
proportional to and ,
and conductance Δ MUT
respectively.
Unlike LC oscillators, the frequency of a ring oscillator
(Fig. 2) is a function of the total load conductance G and
capacitance C of the delay cell (
/
and the
oscillation amplitude is a function of its tail current I and
load conductance
/ . Using the MUT-induced
shifts in oscillation frequency and output amplitude of a
sensor ring oscillator (SVCO) loaded with the sensing
and Δ MUT
can
capacitor and MUT, Δ MUT
and
potentially be determined, and thus
subsequently calculated to fully characterize the MUT
complex permittivity.
In order to avoid precisely measuring the ring oscillator
amplitude to extract distinct capacitance and conductance
shifts, this work proposes a novel two-step procedure that
utilizes an unexposed, MUT-insensitive reference
oscillator and an ALL. Consider first the open-loop case
shown in Fig. 2, with the sensor load substituted with a
dummy capacitor and PMOS resistor Mp in the delay cell
of the reference oscillator (RVCO). After MUT
measurement mode, where
application the first step is
the ALL is activated to set the reference oscillator
amplitude equal to that of the sensor oscillator by
adjusting the Mp gate voltage such that the two delay cells’
conductance are matched. Then the measured frequency
shift between the two oscillators
Δ ALL,ON
can
be
formulated
as
RVCO,ALL,ON
SVCO
Δf ALL,ON ΔCMUT
=
,
f SVCO
C
(1)
which shows that ΔfALL,ON is a function of Δ MUT
only, regardless of MUT loss, and can be used to
measurement mode,
determine . In the second step,
while the MUT is still on top of the sensor the ALL is
Fig. 3. Broadband PLL-based complex dielectric spectroscopy
system.
deactivated and the frequency shift between the two
oscillators Δ ALL,OFF
RVCO,ALL,OFF
SVCO can be
formulated as
Δf ALL,OFF ΔCMUT / C − ΔGMUT / G
=
,
f SVCO
1 + ΔGMUT / G
(2)
which shows that ΔfALL,OFF is a function of both the
and conductance
sensor’s capacitance Δ MUT
. As Δ MUT
has been previously determined
Δ MUT
mode, it is possible to isolate Δ MUT
and
in
determine . In summary, the proposed system only
needs two straightforward frequency shift measurements
to completely and precisely characterize the MUT
complex permittivity.
III. SYSTEM IMPLEMENTATION
Placing the sensing VCO inside an integer-N PLL
allows precise control of the sensing frequency [5], as
shown in the proposed BDS system block diagram (Fig.
3). In parallel, the open-loop reference VCO is controlled
by an RC-filtered version of the PLL control voltage,
Vc,filtered to enable cancellation of frequency drift due to
temperature variations and correlated low frequency noise.
The SVCO and RVCO frequency-vs-voltage curves of
Fig. 4 illustrate the closed-loop system operation. Placing
the MUT on top of SVCO shifts its free running frequency
down by a value that varies with both MUT and .
The PLL then maintains an SVCO frequency equal to Nfref
by adjusting the control voltage from Vc,Air to Vc,MUT, setting
and Δ MUT
. In
GD to compensate for Δ MUT
mode the ALL is enabled to match/increase the RVCO
delay cell conductance to the SVCO value, such that the
frequency shifts upwards relative to SVCO by
Δf ALL,ON
f SVCO
=
f RVCO − Nf ref
Nf ref
=
ΔC MUT
.
C
(3)
control Mp in the reference VCO. The maximum
conductance of Mp is designed to match the maximum
value of Δ MUT , with additional margin to account for
measurement
process variations. Disabling the ALL for
mode is achieved by connecting the gate of Mp to VDD to
maximize its resistance.
IV. EXPERIMENTAL RESULTS
Fig. 4. Sensor and reference VCO frequency versus control
voltage during the MUT characterization procedure.
In
mode the ALL is disabled and the RVCO frequency
returns to its original Vc,MUT value, with a frequency
difference of
Δf ALL,OFF
f SVCO
=
f RVCO − Nf ref
Nf ref
=
ΔCMUT ΔGMUT
.
−
C
G
(4)
Thus, precise frequency shift measurements allow for
and Δ MUT
with Equations
computation of Δ MUT
(3) and (4). Given that the sensing oscillator frequency is
always Nfref, the frequency shifts Δ ALL,ON and Δ ALL,OFF
can be determined by simply measuring the oscillating
frequency of the reference VCO. This is achieved using an
on-chip 32-bit counter to allow for noise filtering by
averaging over a long time interval. This method enables
the same frequency detection precision as in [9], while
also accurately controlling the sample excitation
frequency as in [4], [5].
Aside from the sensor and replica loads, the two ring
oscillators are identical. A three-stage CML topology is
employed to allow for wide-band operation and low
harmonic distortion, with the CML delay cell gain
designed higher than 2 to guarantee oscillation for worstcase MUT loss. Frequency tuning is performed by varying
GD, which consists of a 2-bit discrete resistor bank and a
PMOS transistor in triode for continuous tuning. To make
the oscillation amplitude independent of the PLL control
voltage, the VCO bias current is set via a common mode
feedback circuit that utilizes a replica delay cell and a predefined reference voltage.
In order to form the ALL, two peak detectors are
connected to the reference and sensing VCOs to monitor
the amplitude difference between the two oscillators.
These peak detector outputs feed a high gain opamp to
Fig. 5 shows the chip micrograph of the BDS system,
which is fabricated in a 0.18-µm CMOS process. The
2
system occupies 6.25 mm , and consumes 69-140 mW
from a 1.8V supply. Utilizing a 28.5 MHz reference
frequency, the PLL operates between 0.7-6 GHz for
in the
MUTs with in the range between 1 and 30 and
range between 0 and 30, corresponding to organic
chemical values.
For all the reported results a 300 ms counting time is
used for frequency shift measurements, with each
measurement repeated 10 times and averaged. Measuring
the counter output with the sensor exposed to air allows
for characterization of the system noise, resulting in a
normalized standard deviation of 0.01% with the ALL
OFF and 0.08% with the ALL ON.
System performance over frequency is verified by
exposing the sensor to several binary mixtures of ethanol
and methanol with mixing ratios of methanol q = {0, 20,
50, 80, 100}% and measuring the frequency shifts due to
each mixture with the ALL ON and OFF across 0.76GHz. The mixtures with q = {0, 50, 100}% are utilized
for calibration, as described in [4], [5]. From (3) and (4),
/
and
the relative frequency shifts
,
/
are
functions
of
,
,
/ and Δ MUT
/ , and so they are also
Δ MUT
functions of and , respectively. System calibration is
/
performed with a quadratic equation to fit
,
and
/
as a function
,
,
of the reference mixtures (q = {0,
of reported and
50, 100}%) . For mixtures with q = {20, 80}, and
are determined by substituting frequency shift
measurements in the calibrated equations, with the
Fig. 5. Micrograph of the PLL-based dielectric spectroscopy
chip.
30
TABLE I: PERFORMANCE SUMMARY
25
Theoretical Values
Measurement
ε r'
25
ε r'
20
20
Theoretical Values
Measurement
15
ε
15
''
r
10
10
ε r''
5
5
0
1
2
3
4
5
6
frequency (GHz)
0
0
7
2
x 10
′
and
14
Measurement
Maxwell Garnett
Polder-van Santen f
Coherent Potential
3
4
5
6
frequency (GHz)
9
Fig. 6. Measured and theoretical
mixtures versus frequency.
30
1
′′
7
9
x 10
of ethanol-methanol
Measurement
Maxwell Garnett
Polder-van Santen f
Coherent Potential
12
25
′′
εr
′
εr
10
20
8
15
6
10
5
0
20
40
60
Mixing ratio (q)
80
100
4
0
20
40
60
Mixing ratio (q)
80
100
Fig. 7. Measured and theoretical ′ and ′′ of ethanol-methanol
mixtures versus mixing ratio, q, at different frequencies.
measured and reported values of and
shown in Fig.
6. These measurements show that the maximum error
between the measured and theoretical values over the
entire frequency range is less than 3.7% for both and
.
System performance over mixing ratio is verified using
more binary mixtures of ethanol and methanol with
mixing ratios of methanol q = {0, 10, 20, 40, 50, 60, 80,
100}% for frequencies 0.684, 3.192, 4.56 and 6.156 GHz.
Similar to the previous experiment, mixtures with q = {0,
50, 100}% are utilized for calibration. Fig. 7 shows
measured and theoretical ′ and ′′ versus mixing ratio, q,
with the measurements following the theoretical values
with a maximum error of 3.5%.
Table I summarizes the system performance and
compares this work against other reported CMOS BDS
designs. To the best of our knowledge, this work presents
the first fully integrated system that could characterize
both MUT and . Compared to [6], [7], which require
external frequency synthesizers and voltage-mode ADCs,
the proposed system integrates the sensor inside the onchip VCO/PLL and uses a simple counter for frequency
shift measurements that yield comparable accuracy.
Relative to the LC-VCO system of [4], this work achieves
a much wider 0.7-6 GHz range and is capable of
characterizing both and .
V. CONCLUSION
This work presented a fully integrated BDS system
Specification
This Work
[7]
[6]
[4]
Technology
180 nm
65 nm
350 nm
90 nm
0.05-2.5
7-9
Functionality
,
,
Sensing Frequency
(GHz)
0.7– 6
1-50
Max Error
3.7 %
NA
Power (mW)
69-140
114**
4-9**
16.5
Area (mm2)
6.25
1.2
1.44***
6.25
1% (2GHz)
8.7%(2.5GHz)
3.7 %
* Calculated for |ε|= 4.45 @ 20 GHz
**Does not include frequency synthesizer and ADC.
*** Sensor is not integrated
utilizing a ring oscillator-based PLL for wide band
operation. A novel procedure is proposed for extracting
using two frequency shift measurements,
MUT and
which is considerably simpler than voltage measurement
approaches. The proposed system achieves the highest
CMOS integration level and could accurately characterize
and
of methanol and ethanol mixtures over a wide
frequency range.
ACKNOWLEDGEMENT
This work was supported by the Semiconductor
Research Corporation (SRC) under Task 1836.066
through the Texas Analog Center of Excellence (TxACE).
The authors thank MOSIS for chip fabrication.
REFERENCES
[1] F. Kremer and A. Schonhals, Broadband Dielectric
Spectroscopy, Springer-Verlag, Berlin, Germany, 2003.
[2] Y. Feldman et al., "Time domain dielectric spectroscopy
study of biological systems," TDEI, vol. 10, no. 5, Oct. 2003.
[3] Agilent application note “Basics of measuring the dielectric
properties of materials,” Agilent literature number 59892589EN, August 2013.
[4] A. Helmy et al., "A self-sustained microwave chemical
sensor using a frequency synthesizer," JSSC, vol. 47, no. 10,
Oct. 2012.
[5] O. Elhadidy et al., "A CMOS fractional-N PLL-based
microwave chemical sensor with 1.5% permittivity
accuracy," TMTT, vol. 61, no. 9, Sept. 2013.
[6] M. Bakhshiani et al., "A broadband biosensor interface IC
for miniaturized dielectric spectroscopy from MHz to GHz,"
CICC 2013.
[7] J.-C. Chien et al., "A 1-50 GHz dielectric spectroscopy
biosensor with integrated receiver front-end in 65nm
CMOS," IMS 2013.
[8] M. Bajestan et al., “A 0.62-10GHz CMOS dielectric
spectroscopy system for chemical/biological material
characterization,” to appear in Proc. 2014 IMS.
[9] H. Wang et al., "A frequency-shift CMOS magnetic
biosensor array with single-bead sensitivity and no external
magnet," ISSCC 2009.